ROSS Technology

~ the last words ~

ROSS Technology was incorporated in Texas in August 1988 and spent its whole life as somebody's subsidiary. Cypress Semiconductor bought it in February 1990, at which point it became a Delaware company and Cypress's chip arm in Austin. Fujitsu bought it from Cypress in July 1993. It went public on its own in November 1995 and sold a small piece of itself to Sun Microsystems at the same time, but three years later Fujitsu still held sixty percent of the common stock and all of the preferred, so the float was mostly decoration. Wikipedia credits the founding to Roger D. Ross out of Motorola's microprocessor division, along with several colleagues who left with him, and carries no citation for any of it; the company's own filings never name a founder.

What it sold was a particular kind of cleverness. Sun's workstations of that era took their processors on plug-in modules rather than soldered to the board, and ROSS made the industry's first MBus module in 1990. The hyperSPARC parts that followed were multi-die packages: four to ten separate pieces of silicon, a complete CPU chipset and a tightly coupled second level cache, wired together on a substrate inside one pin grid array so the whole assembly behaved like a single chip. You could open a SPARCstation 20, pull the module out of the socket and put a faster one in, and the machine you had bought three years earlier would run meaningfully quicker. The cache went up to a full megabyte, and clock speeds walked from 100 MHz in late 1994 to 200 MHz parts in 1997.

Sun shipped hyperSPARC in the SPARCstation 20, and Fujitsu, which owned the company, put it in its own DS90 servers. For a while this was a good business to be in, because the installed base of MBus machines was large and the upgrade pitch was easy to make to anyone who had just been told there was no budget for new hardware.

~ the record ~

~ one good year ~

The fiscal year that ended on 1 April 1996 was the only profitable one the company ever had. Net sales of $100.8 million, gross profit of $46.9 million, net income of $18.2 million, $1.23 a share. In the twelve months after that, sales fell to $83.1 million and the company lost $86.7 million. That is a swing of $104.9 million from one fiscal year to the next.

The cause is written in the 10-K in flat language. Sun had introduced a 64-bit architecture called UltraSPARC in 1995, designed in-house, and the filing notes that these products are incompatible with the MBus architecture on which all of the Company's products are based. Sun then moved its entire line to it. Every machine ROSS could sell into was a machine Sun had stopped making, and the total available market for a 32-bit MBus module shrank a little more each quarter without anyone needing to make a decision about ROSS at all.

There was an answer in development, a 64-bit part codenamed Viper. The same filing says it was behind schedule and would not be finished except possibly if the company were sold as a whole, and it lists the Viper team among the people due to be laid off on 31 July 1998 if no buyer appeared. None did.

~ the orderly shutdown ~

Nasdaq wrote on 18 May 1998 to say the stock no longer met the dollar minimum or the public float requirement. On 1 June the board announced what the press release called an orderly shutdown, while still trying to sell what was left. On 31 July, 93 people went, 46 percent of the domestic workforce, and the same day was the last on which anyone could place an end-of-life order. Final shipments had to leave by 28 November.

The manufacturing operation went to BridgePoint Technical Manufacturing for $5.66 million, closing 15 December, after which the company told the SEC it no longer had any operating assets. It also paid BridgePoint a further $1.72 million to take on the worldwide warranty repairs, so part of the sale price went back out of the door to get rid of an obligation. The Israeli design subsidiary went to Fujitsu for $2.5 million. On 29 December 1998 the certificate of dissolution was filed in Delaware and the stock transfer books were closed. Shareholders were told plainly not to expect a distribution of any kind.

~ the last quarter ~

The final 10-Q covers the three months to 28 December 1998 and reads like a receipt. Sales of $4.2 million against $14.3 million a year before, down 70 percent. Nine-month sales of $16.5 million against $37.6 million, and even that number was held up by a single $5.0 million end-of-life order from Fujitsu, the parent buying its own parts one last time before the line stopped. Gross margin 6.7 percent where it had been 39.2.

Nobody filed anything after that. There is no final annual report and no Form 15, because the company dissolved before the paperwork came due and then simply stopped existing, which left its stock registration technically alive and unattended for eighteen years. The SEC finally revoked it on 12 July 2017, describing ROSS Technology in the order as a dissolved Delaware corporation, which it had been since the week after Christmas 1998.

~ the third party in the socket ~

I was fourteen when this company dissolved and the nearest SPARCstation to me was in a building I had no reason to enter. Workstations were not a thing you encountered in Turkey as a teenager with a modem; the machines in the internet cafes a few years later were beige PCs running Counter-Strike, and nobody was swapping anything inside them but the memory. What ROSS sold was a third-party upgrade for somebody else's computer, designed by a company that neither built the machine nor owned the architecture it ran. Desktop processors still come out of sockets. That particular arrangement, where an independent firm sells you a faster version of the chip your vendor shipped, died with the workstation.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »